Show Posts

This section allows you to view all posts made by this member. Note that you can only see posts made in areas you currently have access to.

Messages - Stefan

Pages: 1 ... 129 130 [131] 132 133 ... 164
"AntiPlanet Reflections is released.
It features new 3D engine with full support of reflections.
GPU with CUDA support is required to run AntiPlanet Reflections.
Version with OpenCL support and better CUDA optimization is on the way."

If you are too lazy to play, the game contains about a dozen "system tests".


    * New sensor chips : Fintek F71862, F71869, F71889, F81865, CHIL CHL8266, ITE IT8721, Analog Device ADM1033.
    * Intel QST 2.0.
    * Asus ROG monitoring.
    * Gigabyte DES monitoring.

Go to HWMonitor page for more information.

    * Download HWMonitor 1.16 setup
    * Download HWMonitor 1.16 .zip (32-bit)
    * Download HWMonitor 1.16 .zip (64-bit)

3D-Tech News Around The Web / Blender: Sintel trailer
« on: May 13, 2010, 07:40:53 PM »
"After many weeks of fighting technology, here’s a first glimpse of this wonderful short film that’s shaping up here. With big thanks to Jan Morgenstern for the epic soundtrack, which as usual makes our humble creations look so much better!
We have less than 2 months now to finish this completely… imagine the tension that’s building up here to get everything perfect. For today, we’ll celebrate a big step forward. Enjoy!"

See the announcement on the Durian blog, or download the teaser in a format of your liking or watch on Youtube

This article discusses performance optimizations for AMD GPUs and CPUs using as a case study a simple, yet widely used computationally intensive kernel: Diagonal Sparse Matrix Vector Multiplication. We look at several topics which come up during OpenCL™ performance optimization and apply them to our case study:

   1. Translating C code to OpenCL™
   2. Choosing data structures for dense, aligned memory accesses
   3. Using local, on-chip memory
   4. Vectorizing the computation for higher efficiency
   5. Using OpenCL™ images to improve effective memory bandwidth
   6. Parallelism for multicore processors

At the end of our journey, we'll have a high-performance kernel for both the AMD Radeon™ HD 5870 GPU, as well as the AMD Phenom™ II X4 965 CPU.

Mandelkern - Realtime Mandelbulb in 1k, with some additional effects.

- General:
Has been tested to work on WinXP 32bit with a Geforce 8600M, and WindowsXP 64bit with Radeon 4850. Since it's based on the "Windows 7 fix" version of 1kPack, this demo should be relatively compatible. The 2560x1600 has not been tested because my screen does not support it... Recommended resolution by GPU based on my tests:

Geforce 8600 and similar: 800x600
Geforce 9600/Radeon 3870: 1024x768
Geforce 9800/Radeon 4850: 1280x1024
Geforce GTX/Radeon 4890: 1600x1200/1920x1080
Radeon 58x0: 1920x1200, or even 2560x1600

- Why and what:
Breakpoint 2010 was my second demoparty so far, and even though I was there just as a spectator, I showed my Mandelbulb implementation to a few people, and they suggested I make a 4k demo out of it. Well, I might still finish that 4k eventually (with better performance, more accurate ambient occlusion, blablabla...), but for now here is a 1k demo which raymarches a Mandelbulb and can get decent frame rates at good resolutions on a fast computer. Since the final code was significantly smaller than 1k, I added some additional effects until I reached 1k (well, technically 1k=1000, so it's really 1Ki or 1.024k, but whatever...).

- Algorithm:
The iteration of the Mandelbulb itself is well known, as is the Mandelbrot fractal with orbit traps, so I will just refer to Google for that. The colors are made using a simple Hue-saturation function, which is applied to both the Mandelbulb and the Mandelbrot. The (possibly) new aspect of this visualization here, however, is the tracer, which has a very complex (but still compact) distance estimation function. There are only up to 25 ray marches for each pixel. The (fake) ambient occlusion is based on this tracer, by checking the evolution of the distance function during the tracing.

- Size:
This demo uses hitchhikr's packer and sources, which are essential for making this possible, but there are also a few techniques which help make shader code as small as possible which I discovered during the creation of this demo. Some of those may be obvious, but I will just list all of them here anyway:
- All variables should be just one letter
- Make everything lowercase. The shader in this demo contains not a single uppercase letter.
- All variables should be declared in one place
- Variables should be reused whenever possible in order to minimize the amount of variables which need to be declared (this shader uses 1 float2, 3 float4s and 10 floats)
- Since repetitive strings can be compressed better and some letter are more frequent than others, "1000" may compress better than "999", or "z*z*z*z" may be better than "pow(z,4)"
- Use all features of c++ whenever possible, such as "o=d=1", "while(z<400&&i++<11)" or "x=o-=log(b*b+u*u)*.15"
- DirectX has a few additional features which can be useful. "if(h<1)v=v.gbra" reorders the indices of v while being very compact.
- Truncate floats as much as possible, always ommit the leading zero and remove all unnecessary spaces and brackets.
- Macros are usually not worth it.
- At least on my machine the program would occasionally refuse to start at random. Adding "1*" or "0+" to some line might help.
- Some variables may require an initilization, but the exact value might be not so important. In that case, just initialize them in combination with some other variable ("o=b=a.y;") or reuse them when you know what the last value will look like approximately.

3D-Tech News Around The Web / Alan Wake Technology Trailer
« on: May 13, 2010, 07:02:49 AM »
Microsoft has released a new movie for Alan Wake, the psychological action thriller developed by Remedy Entertainment and scheduled to be available in Europe on May 14th and in the US on May 18th, exclusively on Xbox 360. Alan Wake, the game's protagonist, is a bestselling horror writer, who writes a novel about his darkest nightmares. In the game, those nightmares come true. Players will unravel the riveting plot through multilayered character interactions, unique problem-solving and intense combat against terrifying enemies. Alan Wake features a massive, open world for the players to explore - the game is set in an idyllic all-American small town and its surroundings in the state of Washington. In this nightmarish world, where fantasy and reality collide, Alan Wake must use a variety of weapons to survive, including his most powerful ally against the darkness - light itself.

Download video from GamersHell
Press release: Alan Wake Powered by Havok Physics Technology

"A lot of effort was spent on character lighting" - i hope there will be a PC version before hell freezes over  :-\

Is DirectX 11 Worth Your Time?
Ultimately DirectX 10 will be remembered as a stepping stone. It was a necessary evil that helped rid the DirectX API of over a decade of legacy dead weight. We predict it will be overshadowed in the memories of most gamers by DirectX 11 which looks like it will finally make much of the hype originally created for DirectX 10 a reality. Currently available DirectX 11 titles and hardware really delivers, both in terms of performance and image quality, though arguably not quite at the same time. If you've held out on Vista and DirectX 10, and have been waiting to see if something better comes along before leaving your beloved DirectX 9 WinXP gaming platform behind... well, that something appears to be here and DX11 delivers it.

Full story at HotHW


"When testing cards at below -64°C (sic!) the NVIDIA driver gets confused and thinks the GPU overheats which causes a loss in performance.
When started with /fixcoldbug the attached GPU-Z build will increase the temperature reading that the NVIDIA driver sees by 65°C (it does not actually increase GPU temperature).
This change should result in stable benching at lower temperatures.
Please note that a warm reboot might not remove the change, powering off the system and back on will."

Download and report your success here.

3D-Tech News Around The Web / NVIDIA Cg Toolkit 3.0 Beta available
« on: May 11, 2010, 10:32:47 PM »
Download here

The 3.0 release of Cg contains the following updates:
  New OpenGL GPU Program5 profiles
  New DirectX11 Shader Model 5 profiles
  Support for tessellation programs
  Support for up to 32 texture units
  Unbind routines for D3D programs
  CgFX buffer routines
  Dependent parameter routines for CgFX shader arrays
  Shadow versions of texBLAHproj functions in the hlsl10f profile
  Improved evaluation engine for expressions in CgFX files
  New OpenGL examples including:
  New Direct3D11 examples including;
  Performance improvements and bug fixes

3D-Tech News Around The Web / Games with remarkable physics effects
« on: May 10, 2010, 06:48:20 PM »
"Good physics simulations can make a game feel more realistic and improve the gaming experience - especially if the player benefits from special effects.
Pc Games Hardware takes a look at the past the present and the future of physics in games."

They begin with "Tennis for Two" from 1958...

Download here

Changes from version 2.0.0 B4

  Several interop bugs have been fixed.  If you were having problems with OpenGL or Direct3D
interop, please try this build.
  Documentation for the new interop functions has been added.
  rtDeviceGetName, rtDeviceComputeCapability, rtDeviceGetTotalMemory and
rtDeviceGetAttribute have been unified into a single rtDeviceGetAttribute function.  IMPORTANT! 
Binary compatibility with previous versions of 2.0.0 Beta has been broken.  Binary compatibility
with 1.0 is maintained.
  API functions rtContextGetAttribute, rtContextGetDevices and rtContextGetDeviceCount have
been added.
  Performance improvements to host side node graph processing within OptiX has been improved. 
This should equate to faster frame rates, and should be especially noticeable with scenes
containing many nodes.
  Fixed outstanding bugs related to supporting modifications of the node graph after the first
rtTrace call.
  We no longer depend on the CUDA C runtime library, cudart.  No additional libraries are required
to be distributed with the OptiX libraries at this time.

3D-Tech News Around The Web / Google nukes O3D, supports WebGL
« on: May 10, 2010, 04:21:26 PM »
At Google, we’re deeply committed to implementing and advancing standards, so as of today, the O3D project is changing direction, evolving from its current plug-in implementation into a JavaScript library that runs on top of WebGL.

Full story at Google

Also: Google has acquired BumpTop

UralDev published two comprehensive articles incl. demos (OIT demo didn't run on my rig though).

Use Google for translation if necessary.

3D-Tech News Around The Web / ComputeMark v1.3 available
« on: May 09, 2010, 08:17:38 AM »
Download here

ComputeMark is first 100% DirectX 11 Compute Shader benchmark. With DirectX 11, we can use huge performance of our graphic cards for more (real life, daily used) applications. With ComputeMark we are able to measure this power of our graphic cards and tell which one is better.

ComputeMark typically utilizes 99% of GPU (it's ultimate GPUs ass kicker, great for after-overclocking stability tests) and 0-1% of CPU. Windows Vista or Windows 7, DirectX 11 and DX11 graphic card are required.

- Windows Vista or Windows 7
- DirectX 11
- DX11 graphic card


1.3 (07/05/2010)
- Optimizations for both ATI and NVIDIA
- Specs info on main screen
- Added Windows Vista support

1.2 (17/04/2010)
- Added CrossFire/SLI support

1.1 (12/04/2010)
- Added BURN test

1.0 (10/04/2010)
- First public release

3D-Tech News Around The Web / BlendELF 0.9 Beta released
« on: May 08, 2010, 10:27:58 AM »
BlendELF 0.9 Beta released

Whats new:
- skeletal animations
- C++ and .NET API
- particles ( + mesh emitters )
- light shaft post processing effect
- joystick support
- physics updates (joints, restitution, sleep threshold and limiting physics to axises)
- IPO curves (for animations)
- more tutorials (GUI, particles, C++)
- rendering optimizations
- bug fixes
- Misc.: global tick rate and speed controls, errors recorded to elf.log, scn and gui built in variables removed (functions now return these objects, check the tutorials), over riding F10

Big thanks for Michael Gale for the .NET binding

Check out the new Physics demo with a lot of footballs bouncing around
You can modify amount of balls in init.lua, 5000 balls should be enough to bring your CPU down to the knees  ;D

3D-Tech News Around The Web / Julia 4D demo in 4K using D3D11
« on: May 07, 2010, 04:21:40 AM »
.stem is a Julia 4D demo in 4K using D3D11
Watch video here


Just in case you have missed or ignored the downloads on 3rd party sites, now they are official:


Grass demo has not yet been released  :-\

D3D11/SM5 GPU mandatory, don't try these demos with older GPUs.

"Remedy animation programmer Henrik Enqvist takes a look at how the team created a believable tweed cloth simulation on the title character's jacket in the Microsoft-published Xbox 360 exclusive horror-thriller Alan Wake."

Full story at Gamasutra

Brussels, May 4, 2010 – At the Research@Intel, Europe event in Brussels Intel Labs Europe showed a new technology that could dramatically improve performance of microprocessors when running older software. The technology called “Anaphase” has been developed at Intel Labs Barcelona.

Current processor designs have shifted towards multi-core. Nevertheless single thread performance remains very important as many applications have limited thread-level parallelism. As a result, users executing one of these applications experience little benefit from new multi-core processors.

Researchers from Intel Labs Barcelona now presented “Anaphase” which is a novel hardware/software hybrid approach to leverage multiple cores in order to improve single-thread performance on multi-core processors. This research focuses on different speculative techniques to automatically partition single thread applications to be processed on multiple cores.

On the hardware side, a new unit called “Inter-Core Memory Coherency Module” (ICMC) could be integrated into the die of future processors. The ICMC updates the memory state in program order, detects memory violations and implements check-pointing and recovery mechanisms, so that it can execute the resulting partitioned applications on multiple cores.

First simulated benchmarks running 12 SpecFP and 12 SpecInt benchmarks of the SPEC2006 suite show dramatic average performance improvements of 31% to 41% (depending on core size) over non-Anaphase optimized systems. At the present Anaphase is a research project and the Intel Labs Barcelona researchers are looking into ways how to potentially integrate this technology into future processor designs.


Some whitepapers:

Anaphase: A Fine-Grain Thread Decomposition Scheme for Speculative Multithreading

Microarchitecture and Compilers for Future Processors (TIN2007-61763)

Boosting Single-thread Performance in Multi-core Systems through Fine-Grain Multi-Threading

Pages: 1 ... 129 130 [131] 132 133 ... 164